skip to main content
Explore Scholarly Publications and Datasets in the NSF-PAR

Title: On Occupation Kernels, Liouville Operators, and Dynamic Mode Decomposition
Using the newly introduced ``occupation kernels,'' the present manuscript develops an approach to dynamic mode decomposition (DMD) that treats continuous time dynamics, without discretization, through the Liouville operator. The technical and theoretical differences between Koopman based DMD for discrete time systems and Liouville based DMD for continuous time systems are highlighted, which includes an examination of these operators over several reproducing kernel Hilbert spaces.
Authors:
; ; ;
Award ID(s):
2028001
Publication Date:
NSF-PAR ID:
10281591
Journal Name:
2021 American Control Conference (ACC)
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ( , Journal of Fluid Mechanics)
    Koopman decomposition is a nonlinear generalization of eigen-decomposition, and is being increasingly utilized in the analysis of spatio-temporal dynamics. Well-known techniques such as the dynamic mode decomposition (DMD) and its linear variants provide approximations to the Koopman operator, and have been applied extensively in many fluid dynamic problems. Despite being endowed with a richer dictionary of nonlinear observables, nonlinear variants of the DMD, such as extended/kernel dynamic mode decomposition (EDMD/KDMD) are seldom applied to large-scale problems primarily due to the difficulty of discerning the Koopman-invariant subspace from thousands of resulting Koopman eigenmodes. To address this issue, we propose a framework based on a multi-task feature learning to extract the most informative Koopman-invariant subspace by removing redundant and spurious Koopman triplets. In particular, we develop a pruning procedure that penalizes departure from linear evolution. These algorithms can be viewed as sparsity-promoting extensions of EDMD/KDMD. Furthermore, we extend KDMD to a continuous-time setting and show a relationship between the present algorithm, sparsity-promoting DMD and an empirical criterion from the viewpoint of non-convex optimization. The effectiveness of our algorithm is demonstrated on examples ranging from simple dynamical systems to two-dimensional cylinder wake flows at different Reynolds numbers and a three-dimensional turbulent ship-airwake flow.more »The latter two problems are designed such that very strong nonlinear transients are present, thus requiring an accurate approximation of the Koopman operator. Underlying physical mechanisms are analysed, with an emphasis on characterizing transient dynamics. The results are compared with existing theoretical expositions and numerical approximations.« less
  2. ; ; ; ; ( , Probability Theory and Related Fields)
    Abstract For $$\gamma \in (0,2)$$ γ ∈ ( 0 , 2 ) , we define a weak $$\gamma $$ γ - Liouville quantum gravity ( LQG ) metric to be a function $$h\mapsto D_h$$ h ↦ D h which takes in an instance of the planar Gaussian free field and outputs a metric on the plane satisfying a certain list of natural axioms. We show that these axioms are satisfied for any subsequential limits of Liouville first passage percolation. Such subsequential limits were proven to exist by Ding et al. (Tightness of Liouville first passage percolation for $$\gamma \in (0,2)$$ γ ∈ ( 0 , 2 ) , 2019. ArXiv e-prints, arXiv:1904.08021 ). It is also known that these axioms are satisfied for the $$\sqrt{8/3}$$ 8 / 3 -LQG metric constructed by Miller and Sheffield (2013–2016). For any weak $$\gamma $$ γ -LQG metric, we obtain moment bounds for diameters of sets as well as point-to-point, set-to-set, and point-to-set distances. We also show that any such metric is locally bi-Hölder continuous with respect to the Euclidean metric and compute the optimal Hölder exponents in both directions. Finally, we show that LQG geodesics cannot spend a long time near a straightmore »line or the boundary of a metric ball. These results are used in subsequent work by Gwynne and Miller which proves that the weak $$\gamma $$ γ -LQG metric is unique for each $$\gamma \in (0,2)$$ γ ∈ ( 0 , 2 ) , which in turn gives the uniqueness of the subsequential limit of Liouville first passage percolation. However, most of our results are new even in the special case when $$\gamma =\sqrt{8/3}$$ γ = 8 / 3 .« less
  3. ; ; ; ( , Fluids)
    Learning reservoir flow dynamics is of primary importance in creating robust predictive models for reservoir management including hydraulic fracturing processes. Physics-based models are to a certain extent exact, but they entail heavy computational infrastructure for simulating a wide variety of parameters and production scenarios. Reduced-order models offer computational advantages without compromising solution accuracy, especially if they can assimilate large volumes of production data without having to reconstruct the original model (data-driven models). Dynamic mode decomposition (DMD) entails the extraction of relevant spatial structure (modes) based on data (snapshots) that can be used to predict the behavior of reservoir fluid flow in porous media. In this paper, we will further enhance the application of the DMD, by introducing sparse DMD and local DMD. The former is particularly useful when there is a limited number of sparse measurements as in the case of reservoir simulation, and the latter can improve the accuracy of developed DMD models when the process dynamics show a moving boundary behavior like hydraulic fracturing. For demonstration purposes, we first show the methodology applied to (flow only) single- and two-phase reservoir models using the SPE10 benchmark. Both online and offline processes will be used for evaluation. We observe thatmore »we only require a few DMD modes, which are determined by the sparse DMD structure, to capture the behavior of the reservoir models. Then, we applied the local DMDc for creating a proxy for application in a hydraulic fracturing process. We also assessed the trade-offs between problem size and computational time for each reservoir model. The novelty of our method is the application of sparse DMD and local DMDc, which is a data-driven technique for fast and accurate simulations.« less
  4. ; ; ; ; ; ; ; ; ; ; et al ( , Astronomy & Astrophysics)
    Aims. The dust content of normal galaxies and the dust mass density (DMD) at high- z ( z  > 4) are unconstrained given the source confusion and the sensitivity limitations of previous observations. The ALMA Large Program to INvestigate [CII] at Early times (ALPINE), which targeted 118 ultra-violet (UV)-selected star-forming galaxies at 4.4 <  z  < 5.9, provides a new opportunity to tackle this issue for the first time with a statistically robust dataset. Methods. We exploited the rest-frame far-infrared (FIR) fluxes of 23 galaxies individually detected in their continuum emission, as well as stacked continuum images, to measure the dust content of the 118 UV-selected ALPINE galaxies. We focused on the dust scaling relations and, by comparison with predictions from chemical evolution models, we probed the evolutionary stage of UV-selected galaxies at high- z . By using the observed correlation between the UV luminosity and the dust mass, we estimated the DMD of UV-selected galaxies at z  ∼ 5, weighting the galaxies by means of the UV luminosity function. The derived DMD is compared with the value we estimated from ten ALPINE galaxies blindly detected in the FIR continuum, at the redshift of the ALPINE targets. Results. Our ALMA survey allows the explorationmore »for the first time of the dust content in normal star-forming galaxies at z  > 4 in a statistically robust sample of sources. The comparison of the observed dust scaling relations with chemical evolution models suggests that ALPINE galaxies are not likely progenitors of disc galaxies, but of intermediate- and low-mass proto-spheroids, resulting in present-day bulges of spiral or elliptical galaxies. Interestingly, this conclusion is in line with the independent morphological analysis that shows that the majority (∼70%) of the dust-continuum detected galaxies have a disturbed morphology. The DMD obtained at z  ∼ 5 from UV-selected sources is ∼30% of the value obtained from blind FIR-selected sources, showing that the UV selection misses the most dust-rich, UV-obscured galaxies.« less
  5. ; ( , MRS Advances)
    ABSTRACT In today’s technology, organ transplantation is found very challenging as it is not easy to find the right donor organ in a short period of time. In the last several decades, tissue engineering was rapidly developed to be used as an alternative approach to the organ transplantation. Tissue engineering aims to regenerate the tissues and also organs to help patients who waits for the organ transplantation. Recent research showed that in order to regenerate the tissues, cells must be seeded onto the 3D artificial laboratory fabricated matrices called scaffolds. If cells show healthy growth within the scaffolds, they can be implanted to the injured tissue to do the regeneration. One of the biggest limitation that reduces the success rate of tissue regeneration is the fabrication of accurate thick 3D scaffolds. In this research “maskless photolithography” was used to fabricate the scaffolds. Experiment setup consist of digital micro-mirror devices (DMD) (Texas Instruments, DLi4120), optical lens sets, UV light source (DYMAX, BlueWave 200) and PEGDA which is a liquid form photo-curable solution. In this method, scaffolds are fabricated in layer-by-layer fashion to control the interior architecture of the scaffolds. Working principles of the maskless photolithography is, first layer shape is designedmore »with AutoCAD and the designed image is uploaded to the DMD as a bitmap file. DMD consists of hundreds of tiny micro-mirrors. When the UV light is turned on and irradiated the DMD, depending on the micro-mirrors’ angles, UV light is selectively reflected to the low percentage Polyethylene (glycol) Diacrylate (PEGDA) photo-curable solution. When UV light penetrates into the PEGDA, only the illuminated part is solidified and non-illuminated area still remains in the liquid phase. In this research, scaffolds were fabricated in three layers. First layer and the last layer are solid layers and y-shape open structure was sandwiched between these layers. After the first layer is fabricated with DMD, a “y-shape” structure was fabricated with the 3D printer by using the dissolvable filament. Then, it was placed onto the first solid layer and covered with fresh high percentage PEGDA solution. UV light was reflected to the PEGDA solution and solidified to make the second and third layers. After the scaffold was fabricated, it is dipped into the limonene solution to dissolve the y-shape away. Our results show that thick scaffolds can be fabricated in layer-by-layer fashion with “maskless photolithography”. Since the UV light is stable and does not move onto the PEGDA, entire scaffold can be fabricated in one single UV shot which makes the process faster than the current fabrication techniques.« less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email